SoftBank’s AI power bet led to a $7.1 billion loss

More than half of SoftBank’s quarterly loss was tied to turbine deposits and power equipment, exposing the cost of racing to secure energy for AI.

SoftBank’s push into natural gas power for AI data centers has turned into an expensive setback. The company reported a $7.1 billion quarterly loss on Tuesday, with more than half of it tied to losses on power-generating equipment. The result puts a spotlight on a less visible part of the AI boom — the scramble to lock in enough electricity.

That matters well beyond SoftBank. As more data centers come online, power is becoming as critical as chips and software, and companies are making big, risky commitments to secure supply. According to IT-PUB News, SoftBank’s latest results show how costly that strategy can get when demand expectations, procurement timing, and market conditions drift out of sync.

SoftBank ordered more turbines than it needed

According to the source, SoftBank started placing deposits on more than 13 gigawatts of turbines about two years ago, even though that was nearly 50% above its needs. The over-ordering reportedly helped the company secure turbines and move projects ahead faster.

The move also fits SoftBank’s broader habit of making large bets when it sees an opening. Earlier in 2025, it committed $30 billion to OpenAI, and in February it said its SB Energy subsidiary would spend $33 billion on a 9.2-gigawatt natural gas power plant.

The idea was simple enough: if AI companies were going to need power quickly, SoftBank wanted to be in position to supply it. But ordering turbines at that scale also left the company exposed if demand shifted or project schedules changed.

Turbine scarcity made the strategy more costly

The energy market was not prepared for how fast AI’s electricity demand would grow. Turbine delivery backlogs now stretch into the early 2030s, making equipment harder to get and pushing buyers to put down large deposits far in advance.

Peter Gardett, CEO of Noreva, told TechCrunch by email that SB Energy “did not have a choice except to overbook turbines in the future if they wanted some turbines right away.” In that view, scarcity shaped SoftBank’s strategy as much as ambition did.

But the same decision that helped secure supply created a problem once the company had to account for the deposits. SoftBank had paid turbine manufacturers nonrefundable deposits and now considered much of that money lost. Its quarterly report described a charge of more than $3 billion tied to that miscalculation.

Gardett said the loss reflects “price discovery” finally catching up with the market. Instead of risking even higher costs later and still coming up short, SoftBank over-ordered and overpaid early.

Natural gas remains part of the AI buildout

SoftBank’s loss does not mean the AI industry has stopped looking for new power sources. The source says electricity demand from AI is still too large for natural gas to drop out of the picture. For now, gas is still “riding high,” even if SoftBank may have run into the limits of how far that approach can be pushed.

At the same time, some data center operators appear to be growing more cautious. As they get more experience running AI facilities, they are gaining a clearer sense of what is actually needed to keep them online. Natural gas may be familiar and dependable, but it still comes with risks.

Brent Nelson, senior managing director of markets and strategy at Ascend Analytics, told the Equity podcast Wednesday that winter curtailments remain a concern when supplies tighten and demand rises. In other words, gas can come under pressure at exactly the moment operators need it most.

Nelson said the grid may still be the better option for many data centers, even if they ask for a smaller connection than usual. He also said batteries could help cover the gap, especially for shaving peak loads or keeping operations running if a utility asks them to cut electricity use.

SoftBank’s loss could make other companies more careful

The company’s setback is unlikely to cause a broad retreat from gas-fired power projects. The source says AI’s power demand remains too large for that. Still, SoftBank’s results may push other companies to think harder about how quickly they lock themselves into large energy deals.

That is the larger point here. The AI boom is not just about models and data centers. It is also about the infrastructure underneath them, and the cost of securing that infrastructure is rising fast. SoftBank moved early and at scale, and now the price of that decision is out in the open.

For businesses building AI systems, power is no longer a background concern. It is becoming a strategic asset — and an expensive one. SoftBank’s quarterly loss shows how easily a rush to meet AI’s energy needs can turn into a costly misstep.